rio, lca, refactoring

[jalview.git] / forester / java / src / org / forester / util / SequenceIdParser.java

diff --git a/forester/java/src/org/forester/util/SequenceIdParser.java b/forester/java/src/org/forester/util/SequenceIdParser.java
index 7486d8a..3b6bc5d 100644 (file)
--- a/forester/java/src/org/forester/util/SequenceIdParser.java
+++ b/forester/java/src/org/forester/util/SequenceIdParser.java
@@ -32,12 +32,9 @@ import java.util.regex.Matcher;
 import java.util.regex.Pattern;\r
 \r
 import org.forester.phylogeny.data.Identifier;\r
-import org.forester.ws.uniprot.DatabaseTools;\r
 \r
 public final class SequenceIdParser {\r
 \r
-    \r
-   \r
     // gb_ADF31344_1_segmented_worms_\r
     // gb_AAA96518_1\r
     // gb_EHB07727_1_rodents_\r
@@ -47,7 +44,6 @@ public final class SequenceIdParser {
     // mites|ref_XP_002434188_1\r
     // ref_XP_002434188_1_mites___ticks_\r
     // ref_NP_001121530_1_frogs___toads_\r
-    \r
     //The format for GenBank Accession numbers are:\r
     //Nucleotide: 1 letter + 5 numerals OR 2 letters + 6 numerals\r
     //Protein:    3 letters + 5 numerals\r
@@ -58,27 +54,98 @@ public final class SequenceIdParser {
                                                                          .compile( "(?:\\A|.*[^a-zA-Z0-9])([A-Z]{2}\\d{6})(?:[^a-zA-Z0-9]|\\Z)" );\r
     private final static Pattern GENBANK_PROTEIN_AC_PATTERN      = Pattern\r
                                                                          .compile( "(?:\\A|.*[^a-zA-Z0-9])([A-Z]{3}\\d{5})(?:[^a-zA-Z0-9]|\\Z)" );\r
-    private final static boolean DEBUG                           = false;\r
+    // RefSeq accession numbers can be distinguished from GenBank accessions \r
+    // by their distinct prefix format of 2 characters followed by an\r
+    // underscore character ('_'). For example, a RefSeq protein accession is NP_015325. \r
+    private final static Pattern REFSEQ_PATTERN                  = Pattern\r
+                                                                         .compile( "(?:\\A|.*[^a-zA-Z0-9])([A-Z]{2}_\\d{6,})(?:[^a-zA-Z0-9]|\\Z)" );\r
+    // See: http://web.expasy.org/docs/userman.html#ID_line\r
+    private final static Pattern TREMBL_PATTERN                  = Pattern\r
+                                                                         .compile( "(?:\\A|.*[^a-zA-Z0-9])([A-Z][0-9][A-Z0-9]{3}[0-9])(?:[^a-zA-Z0-9]|\\Z)" );\r
 \r
-    \r
-    \r
+    /**\r
+     * Returns null if no match.\r
+     * \r
+     */\r
     public final static Identifier parse( final String s ) {\r
-        String v = DatabaseTools.parseGenbankAccessor( s );\r
+        String v = parseGenbankAccessor( s );\r
         if ( !ForesterUtil.isEmpty( v ) ) {\r
-            return new Identifier( v, "genbank" );\r
+            return new Identifier( v, Identifier.NCBI );\r
+        }\r
+        v = parseRefSeqAccessor( s );\r
+        if ( !ForesterUtil.isEmpty( v ) ) {\r
+            return new Identifier( v, Identifier.REFSEQ );\r
+        }\r
+        v = parseTrEMBLAccessor( s );\r
+        if ( !ForesterUtil.isEmpty( v ) ) {\r
+            return new Identifier( v, Identifier.SP );\r
         }\r
-        \r
         return null;\r
     }\r
-    \r
-    \r
-    \r
-    \r
-    \r
+\r
+    public final static boolean isProtein( final String query ) {\r
+        final String r1 = parseRefSeqAccessor( query );\r
+        if ( !ForesterUtil.isEmpty( r1 ) && ( r1.charAt( 1 ) == 'P' ) ) {\r
+            return true;\r
+        }\r
+        final String r2 = parseTrEMBLAccessor( query );\r
+        if ( !ForesterUtil.isEmpty( r2 ) ) {\r
+            return true;\r
+        }\r
+        return GENBANK_PROTEIN_AC_PATTERN.matcher( query ).lookingAt();\r
+    }\r
+\r
+    /**\r
+     * Returns null if no match.\r
+     * \r
+     */\r
+    public static String parseGenbankAccessor( final String query ) {\r
+        Matcher m = GENBANK_NUCLEOTIDE_AC_PATTERN_1.matcher( query );\r
+        if ( m.lookingAt() ) {\r
+            return m.group( 1 );\r
+        }\r
+        else {\r
+            m = GENBANK_NUCLEOTIDE_AC_PATTERN_2.matcher( query );\r
+            if ( m.lookingAt() ) {\r
+                return m.group( 1 );\r
+            }\r
+            else {\r
+                m = GENBANK_PROTEIN_AC_PATTERN.matcher( query );\r
+                if ( m.lookingAt() ) {\r
+                    return m.group( 1 );\r
+                }\r
+                else {\r
+                    return null;\r
+                }\r
+            }\r
+        }\r
+    }\r
+\r
+    /**\r
+     * Returns null if no match.\r
+     * \r
+     */\r
+    private final static String parseRefSeqAccessor( final String query ) {\r
+        final Matcher m = REFSEQ_PATTERN.matcher( query );\r
+        if ( m.lookingAt() ) {\r
+            return m.group( 1 );\r
+        }\r
+        return null;\r
+    }\r
+\r
+    /**\r
+     * Returns null if no match.\r
+     * \r
+     */\r
+    private final static String parseTrEMBLAccessor( final String query ) {\r
+        final Matcher m = TREMBL_PATTERN.matcher( query );\r
+        if ( m.lookingAt() ) {\r
+            return m.group( 1 );\r
+        }\r
+        return null;\r
+    }\r
+\r
     private SequenceIdParser() {\r
         // Hiding the constructor.\r
     }\r
-    \r
-    \r
-    \r
 }\r